StarCrack AI Official Feedback Thread (V5.5)

[rackdude]

Stats for probes:


12Probes: 457-533
16Probes: 629-686
20Probes: 725-781
22Probes: 743-819
+24Probes: 762-838+
26Probes: 762-838
30Probes: 781-857
34Probes: 781-857
[/size][/color]
[/size][/color]You can 150 more per minute with the extra probes. Besides, by 18 probes you should have an expansion up.
[/size][/color]
[/size][/color]The AI pushes should come when (as Protoss) you have around 140 supply. They should fall simply. After that you should be at 200/200 for until you clear them out. If you are ever below 150 after their first push, be looking at your macro. I've found I could get 2 expansions by the first push and then when I push out to kill the first comp I take a 4th base (on Lost Temple the close island). After that I sweep to the other comp and take the last comps Nat. By then it's just too many minerals to use, massive non-stop 200/200 with units just warping the moment they die right into the battle field (if they die). The AI doesn't even stand a chance at that point.
[/size][/color]
[/size][/color]So my suggestion: the AI should be at 200/200 mid/late game. I don't care if it has to cheat to do so, but now that I'm at least noobie instead of ultra-noobie at macro, I notice how much quicker I can get to 200/200 than the cheating AI.

[Nesreca]

Btw, is it only me that can't wait to test my noob skills vs real person ?

You're not the only one. Keep working on your 1337 skills though, there are some pretty crazy good players out there that will totally kick asses.

[Crixomix]

Not to sound ungrateful, but when should we expect a new AI? (I'm assuming version 6.0)
There was a new version or two every day for a while, now it's been two days since an update. I'm dyin here ><


To be honest version 3, 4, and 5, felt pretty similar to me. Except each one gets slightly better, such as more units and better structuring of their base, but they don't feel any smarter than before.



Granted, you guys are doing an AMAZING job making these AIs. And I am very thankful. I'm hoping for some big changes in 6.0.
Keep truckin

[cliffton]

they dont saturate mineral fields (builds 15 scvs and stops for a good while - saturation is at 24)

24 is saturation? I only go with 16 and I fare pretty well. I don't see a need for 24, that's just overkill. It'd only be good if the extra 8 were going for expansion, and if they don't, that's a waste of 400 minerals.

I agree. At the closest mineral (because they do stagger slightly in distance), 2 SCVs is fully saturated. So technically it's more like 2-2.5 SCVs per mineral for saturation, but to me that's not fully necessary. So 16 SCVs per standard mineral batch (12 for the high yield) is a nearly full saturation. 24 is definitely overkill.

24 is certainly not overkill.  That's the number of probes that gives you the fastest income.  Besides, you are going to be expanding at some point.  When the new nexus goes up, you're going to want to move some of your probes from your main to your expansion (maynarding) so it is up and running as fast as possible without having to build as many new probes from the new nexus.  Keeping that in mind I will often oversaturate, then when a new expansion comes up it is mining at a fast rate right away

[Gamewiz]

they dont saturate mineral fields (builds 15 scvs and stops for a good while - saturation is at 24)

24 is saturation? I only go with 16 and I fare pretty well. I don't see a need for 24, that's just overkill. It'd only be good if the extra 8 were going for expansion, and if they don't, that's a waste of 400 minerals.

I agree. At the closest mineral (because they do stagger slightly in distance), 2 SCVs is fully saturated. So technically it's more like 2-2.5 SCVs per mineral for saturation, but to me that's not fully necessary. So 16 SCVs per standard mineral batch (12 for the high yield) is a nearly full saturation. 24 is definitely overkill.

24 is certainly not overkill.  That's the number of probes that gives you the fastest income.  Besides, you are going to be expanding at some point.  When the new nexus goes up, you're going to want to move some of your probes from your main to your expansion (maynarding) so it is up and running as fast as possible without having to build as many new probes from the new nexus.  Keeping that in mind I will often oversaturate, then when a new expansion comes up it is mining at a fast rate right away

Makes sense. I'll give it a try next time I play.

[Gamewiz]

I know this has been mentioned, but I want to stress this as I think it's very important. Please please please make the AI destroy rocks and things to get to other areas or expansions. Places like Kulas Ravine are so easily exploitable because the AI will *not* destroy rocks, which means those side expansions will never be accessed by them, and they will never try to destroy the rocks that allow shorter passageway from left to right.

[creamsoda]

24 is saturation? I only go with 16 and I fare pretty well. I don't seea need for 24, that's just overkill. It'd only be good if the extra 8were going for expansion, and if they don't, that's a waste of 400minerals.

If you're new to starcraft, I suggest you listen to the advice from the more experienced SC1 players.. 24 workers is definitely NOT overkill...

[Swarmling]

ive found out with the AI if u were to trap there units and your army is bigger then theres they just try and run by your units and not even attack but they wont even make it they just all die so to make the game fun when they run i run  they only really attack if they know there going to win the battle or your attacking there main base so they try to do anything to stop you funny how the works run around the base all crazy like shouldnt they all run to there other base and mine or somthing but the AI is really good and fun
they just lack some micro and never use many spells

[JPeterson]

I've made some changes to the terran and toss late game to make it harder. You can try it from the latest svn here. I removed a bottleneck for terran that made it be stuck on vikings in the late game. And I added a new ground units late game for toss where it will use colossus and observers. I also made sure toss and terran had transportation in the late game so that they can expand to the islands, or high ground like in Shakuras Plateau, and attack your island expansions.

As usual if you have a decent APM you can kill the AI any time. But this should have terran attack with bigger groups of vikings, thors and tanks in the late game. And have toss attack with bigger groups of zealots, stalkers, immortals, colossus and observers. So that they become little more of a challenge.

[Buzzard]

I just tried this AI out on two maps that are listed as AI maps and these are: Metalopolis (6 players) and Kulas Ravine (4 players), as my favorite race is Zerg, i easily won both times.

It seems that the AI is too much of a tech builder and does not take out any units in the early game, instaid it saves for heavy cost units that are totally worthless in a 15min early game. It also do not rush or does any early attack to prevent any incoming attack from other ai or players. So im hoping this can be improved.

To me this is almost like the AI that blizzard made in SC1. Anyway, good luck!.

[JPeterson]

It seems that the AI is too much of a tech builder and does not take out any units in the early game, instaid it saves for heavy cost units that are totally worthless in a 15min early game.

Which race did you play against?

[Buzzard]

It seems that the AI is too much of a tech builder and does not take out any units in the early game, instaid it saves for heavy cost units that are totally worthless in a 15min early game.

Which race did you play against?

Protoss, terran and zerg.

Zerg went straight for mutalisks and by the time they got out one, i overran thier base with hydralisks

Protoss..well.. they went for air units as well and by the time they got out some, i overran them.

Terran, i think they went for mech/bio build but still, same there.

[itsarabbit]

I've made some changes to the terran and toss late game to make it harder. You can try it from the latest svn here. I removed a bottleneck for terran that made it be stuck on vikings in the late game. And I added a new ground units late game for toss where it will use colossus and observers. I also made sure toss and terran had transportation in the late game so that they can expand to the islands, or high ground like in Shakuras Plateau, and attack your island expansions.

As usual if you have a decent APM you can kill the AI any time. But this should have terran attack with bigger groups of vikings, thors and tanks in the late game. And have toss attack with bigger groups of zealots, stalkers, immortals, colossus and observers. So that they become little more of a challenge.

Will it be included in the official starcrack 2 AI release?
Also, which guys does the AI team consist of?

[Tijnnn]

I know this has been mentioned, but I want to stress this as I think it's very important. Please please please make the AI destroy rocks and things to get to other areas or expansions. Places like Kulas Ravine are so easily exploitable because the AI will *not* destroy rocks, which means those side expansions will never be accessed by them, and they will never try to destroy the rocks that allow shorter passageway from left to right.

If you want it to make little harder, so detroy the rocks yourself with some range units.

I really like AI 5.5 something to kill atleast, zerg is fast rusher, terran and protoss are bit slower. Protoss normaly builds mass carriers in l8game. And the Terran builds marines and marauders at begin.
When i play a random game i get everything double'' main building + builders. anyone knows how to fix it? i think i got a file 2 times somewhere. dont wanna go in there and delete stuff i need.

Keep up the good!thx

[itsarabbit]

maybe 6.5 or 7.0 can have some cheese strategies?
cheese = builds made to surprise the enemy, like building raxes right outside base, very hard(impossible?) with zerg, as they need creep to build stuff.